Professional Documents
Culture Documents
SCOPE
A significant fraction of chemical process design is con- alize Guggenheims method, UNIQUAC contains no more
cerned with separation of fluid mixtures by diffusional than two adjustable parameters per binary.
operations. All design methods for such separations re- UNIQUAC is applicable to multicomponent mixtures
quire quantitative estimates of fluid-phase equilibria; of nonpolar and polar liquids (including those that partici-
this work provides a contribution toward making such pate in hydrogen bonding) as encountered in typical chem-
estimates for liquid-phase mixtures of nonelectrolytes, in- ical and petrochemical processes. No ternary (or higher)
cluding polymers, and including those mixtures where constants are required. Attention is given to vapor-liquid
nonideality is sufficiently strong to produce two liquid and liquid-liquid equilibria.
phases. When well-defined simplifying assumptions are made,
Activity coefficients in liquid mixtures can be calculated the UNIQUAC model can yield any one of several well-
from a model which expresses the excess Gibbs energy known expressions for the excess Gibbs energy, including
of the mixture as a function of the composition. A new the van Law, Wilson, and NRTL equations. Relative to
model, called UNIQUAC, is presented here. This model these well-known equations, the advantage of UNIQUAC
is derived from a statistical-mechanical basis extending is that, for a large variety of multicomponent systems and
that used by Guggenheim in his quasi-chemical theory. using only two adjustable parameters per binary, reliable
Unlike Guggenheims theory, however, UNIQUAC is estimates can be made of both vapor-liquid and liquid-
applicable to mixtures wh0.e molecules differ appreciably liquid equilibria using the same equation for the excess
in size and shape and, unlike previous attempts to gener- Gibbs energy.
The thermodynamics of nonideal liquid mixtures has In this work, the theory of Guggenheim is extended to
generated a vast literature. Despite attention for over a mixtures containing molecules of different size and shape
century from some of the best scientific minds, the goal by utilizing the local-composition concept introduced by
of predicting mixture properties from pure-component Wilson (1964). The central idea of this concept is that
properties alone remains elusive; failure to reach this goal when viewed microscopically, a liquid mixture is not
follows from inadequate fundamental understanding of homogeneous; the composition at one point in the mixture
liquid structure and intermolecular forces. Therefore, to cal- is not necessarily the same as that at another point. In
culate vapor-liquid and liquid-liquid equilibria needed for engineering applications and in typical laboratory work
process design, it is necessary to construct models which, only the average, overall (stoichiometric) composition
by necessity, are only approximations containing param- matters, but for constructing liquid-mixture models, it
eters that must be obtained empirically. This work pre- appears that the local composition, rather than the average
sents such a model and discusses its use for engineering composition, is a more realistic primary variable.
applications. The extension, or generalization, of Guggenheims
model leads to a result here called the UNIQUAC (uni-
ACTIVITY COEFFICIENT, EXCESS GIBBS ENERGY AND versal quasi-chemical) equation. The derivation of this
LATTICE MODELS equation is given in the next few sections. Readers con-
cerned only with the engineering application of this equa-
Deviations from ideal behavior (Raoults law) are tion may proceed directly to the section on Application to
commonly expressed by activity coefficients as discussed Binary and Multicomponent Systems following Equation
in numerous textbooks. In a mixture, activity coefficient
(24).
yi (for component i) is related to gE, the excess Gibbs
energy per mole of mixture, by Partition Function for a Binary Liquid Mixture
nTgE= RT I; ni In yi (1) Following Guggenheim (1952), we postulate that a
i liquid can be represented by a three-dimensional lattice
of equi-spaced lattice sites; the volume in the immediate
vicinity of a site is called a cell. Each molecule in the
liquid is divided into attached segments such that each
segment occupies one cell. The total number of cells is
where ni is the number of moles of component i and nT equal to the total number of segments. [A possible refine-
is the total number of moles. To obtain activity coefficients, ment where some cells are unoccupied (holes) is not used
therefore, it is necessary to construct an expression which here.] The configurational partition function 2 is given by
gives g E as a function of composition, temperature, and
pressure; the most important variable is composition. For 2 = Zlattice * &ell (3)
liquid mixtures at ordinary pressures, the effect of pres-
sure is negligible. The effect of temperature is not negligi- where Zlatticerefers to the situation where the center of
ble, but often it is not large when consideration is re- every segment is coincident with a lattice site and where
stricted to a moderate temperature range. Zcell provides those contributions to Z which are caused
When nonelectrolyte liquids are mixed at constant tem- by motions of a segment about this central position. In
perature and constant pressure remote from critical con- mixtures of nonelectrolyte liquids removed from critical
ditions, there is little volume change. As shown by Scatch- conditions, we assume that for each component &ell is
ard (1937), even small volume changes can have a independent of composition. For a binary mixture contain-
significant effect on the entropy of mixing and on the ing N1 molecules of component 1 and N 2 molecules of
enthalpy of mixing but, to a good approximation, these component 2, the Helmholtz energy of mixing is then
effects tend to cancel in the excess Gibbs energy. There- given by
fore, when attention is restricted to mixtures of nonelectro-
lyte liquids at low or modest pressures, we can substitute
for the excess Gibbs energy of mixing at constant tempera-
ture and pressure, the excess Helmholtz energy of mixing where lc is Boltzmanns constant. The molar excess Gibbs
at constant temperature and volume (Hildebrand and energy g is given by
Scott, 1950). This substitution very much facilitates con-
struction of a theory which is physically reasonable on the
one hand and mathematically (relatively) simple on the
other. Such a theory is provided by Guggenheims quasi-
chemical lattice model which, however, in its original form where R is the gas constant, x stands for mole fraction, and
is restricted to small molecules of essentially the same size. n for the number of moles.
Previous attempts to extend it to larger molecules (Barker, Following Guggenheim, the lattice partition function is
1952; Sweeny and Rose, 1963) have not been useful, pri- given by
marily because of the excessive number of adjustable
parameters required to reduce the theory to practice. Zlattice = 8 4 6 ) exp [ - UO(~)/~TI (6)
a11e
where ~ ( 0 is
) the combinatorial factor given by Staverman
( 1950).
Having found h as outlined above, we now proceed to
find the next approximation for the average local fractions and
for the nonathermal case, that is, where uij # 0.
gE (residual)
AVERAGE LOCAL AREA FRACTIONS IN NONATHERMAL
= -41x1 in 181 + e2 TZ11
RT
MIXTURES
The summation in Equation (6) is replaced by its
maximum term, Equations ( l o ) , (11), and (12) are used
again but in this approximation U O [Equation (9)] is not
set equal to zero. Equation (17) is retained. The resulting
expression for Zlattice is separately differentiated with re-
spect to ell and eZ2 and the results are set equal to zero.
Again using the constraining Equations (7) and (8) but
omitting mathematical details, we now find that the aver- In Equation (21) @ is the average segment fraction:
age local area fractions are given by
ell(l) = el (18)
- (u21 - u11)
81 + 4 exp RT
Equations (20) to (22) contain pure-component struc-
tural parameters rl, r2, 41, and q2; these are evaluated
and from bond angles and bond distances as discussed in Ap-
022 (1) = 82 pendix B.
(19) Note that the expression for gE (combinatorial) contains
- (u12 - u22)
e2 + o1exp two composition variables: the average area fraction 8
RT and the average segment fraction @. However, the ex-
where Uij is expressed in units of calories per mole and pression gE (residual) contains only one composition vari-
where superscript ( 1 ) denotes first approximation and el able: the average area fraction 8. There are no adjustable
and 82 are average area fractions defined by Equations binary parameters in Equation (21), but there are two
(15) and (16). adjustable binary parameters in Equation (22) : (uz1 -
In mixtures that are not athermal, therefore, the average u11) and (u12 - ~ 2 2 ) . From the derivation of Equations
local area fractions are not the same as the average area (20) to (22) it follows that uzl = u12.
fractions. Relations similar to Equations (18) and (19) APPLICATION TO BINARY AND
were previously suggested by Wilson ( 1964). MULTICOMPONENT SYSTEMS
Equations (20) to (22) give the excess Gibbs energy
for a binary mixture in terms of two adjustable binary
VALUFSOF SIZEAND SURFACE
TABLE1. TYPICAL parameters and two pure-component structural parameters
PARAMETERS* per component, r and q. Table 1 gives these structural
Fluid r 4 parameters for some representative nonelectrolyte mole-
cules; a more complete list is available from the authors
Water 0.92 1.40 upon request.
Carbon dioxide 1.30 1.12
Acetaldehyde 1.90
Activity coefficients are readily found by differentiation
1.80
Ethane 1.80 1.70 as indicated by Equation ( 2 ) . For a binary mixture, activ-
Dimethyl amine 2.33 2.09 ity coefficient 71 is given by
Methyl acetate 2.80 2.58
Furfural 2.80 2.58
Benzene 3.19 2.40
Toluene 3.87 2.93
Aniline 3.72 2.83
Triethyl amine 5.01 4.26
n-Octane 5.84 4.93
n-Decane 7.20 6.02
n-Hexadecane 11.24 9.26
Acetone 2.57 2.34
Chlorofo m 2.87 2.41
* Parameters are evaluated as shown in Appendix B.
+ -8qixiln-
2 $1
(2la)
2. SIMPLIFICATIONS
TABLE IN UNIQUAC EQUATIONTO
2 i @pi
OBTAINWELL-KNOWN RELATIONSFOR THE EXCESS
gE (residual) GIBBSENERCY
= - ?qix+ln
RT
88jTji
(j 1 (22a)
Simplifying assumptions
Equation Resulting relation
(21) Equation (22) for g E ( a )
(3
truncating after 91 = ci
quadratic term 92 =02
Equals zero Expand terms in Three-suffix Margules
TABLE3. TYPICAL
REPRESENTATION VAPOR-LIQUID
OF BINARY BY UNIQUAC EQUATION
EQUILIBRIA
42,Variance of Fit x 1031
Temp., "C or No. of data One Two
System pressure, mm Hg points parameter parameter Reference
Methylcyclopeatane-benzene 760 29 8.4 8.7 Myers ( 1956)
Benzene-iso-octane 760 11 4.1 4.5 Chu (1956)
Isooctane-nitroethane 35" 19 0.4 0.2 Edwards (1962)
Nitromethane-benzene 45" 12 0.4 0.5 Brown (1957)
Hexane-nitroethane 45" 12 1.3 0.1 Edwards (1962)
Methanol-benzene 55" 9 197.2 3.2 Scatchard (1946)
Ethanol-iso-octane 50" 13 103.4 6.3 Kretschmer ( 1948)
Ethanol-hexane 7430 16 196.3 8.8 Sinor (1960)
Ethanol-water 70" 13 126.1 0.55 Mertl ( 1972)
Ethylacetate-ethanol 70" 15 18.2 0.69 Mertl ( 1972)
Water-methylethyl ketone 760 8 1.7 1.3 Othmer (1945)
Acetone-benzene 45" 11 2.7 0.5 Brown ( 1957)
Carbon tetrachloride-acetonitrile 45" 13 147.8 7.1 Brown (1954)
Methylaceta te-ethanol 45" 11 8.6 9.3 Nagata (1962)
Acetone-chloroform 50" 29 23.5 4.2 Severns ( 1953)
Methylacetate-methanol 50" 15 41.6 4.2 Severns (1953)
t The variance of the fit @ is given by
where the superscript M denotes the measured value of the variable and the superscript o denotes the estimate of the true value of the variable. D
is the number of data points and L is the number of parameters and the 0's are the variances in the measured variables.
TABLE4. BINARYPARAMETERS
IN TWO-PARAMETER UNIQUAC EQUATION
OR ONE-PARAMETER FOR REPRESENTATIVE
BINARYSYSTEMS
Temp., "C or Two-parameter One
pressure, (u21 - u11) (u12 - un) parameter
System ( 1 ) - ( 2 ) mm Hg cal/mole ca1/mo1e (712 Reference
Methylcyclopentane- 760 -36.9 138.1 0.092 Myers (1956)
benzene
Benzene-iso-octane 760 182.1 -76.5 0.078 Chu ( 1956)
Isooctane-nitroethane 35" 5.3 492.3 0.412 Edwards (1962)
Nitromethane-benzene 45" 309.1 35.45 0.218 Brown (1957)
Hexane-nitroethane 45" -36.3 471.6 0.351 Edwards ( 1962)
Methanol-benzene 55" 1355.8 -417.4 0.182 Scatchard (1946)
Ethanol-iso-octane 50" 968.2 -357.6 0.229 Kretschmer (1948)
Ethanol-hexane 760 940.9 -335.0 0.308 Sinor (1960)
Water-ethanol 70" 258.4 378.1 0.272 Mertl (1972)
Ethylacetate-ethanol 70" -292.3 446.5 0.034 Mertl (1972)
Water-methylethyl ketone 760 622.3 222.2 0.419 Othmer (1915)
Acetone-benzene 45" 331.0 -208.9 0.059 Brown (1957)
Carbon tetrachloride- 45" - 100.1 953.4 0.494 Brown (1954)
acetonitrile
Methylacetate-ethanol 45" -40.5 426.5 0.229 Nagata (1962)
Acetone-chloroform 50" 149.8 -315.5 -0.183 Sevems ( 1955)
Methylacetate-methanol 50" -233.1 622.1 0.156 Severns (1955)
POLYMER SOLUTIONS
I I I I I i Since the combinatorial contribution to the excess Gibbs
I 0
0
L i q u i d - L i q u i d Doto
V o p o r - L i q u i d Doto I energy [Equation (21)] is applicable to mixtures contain-
ing very large molecules (polymers) and molecules of nor-
mal size (monomers), the UNIQUAC equation may also
be used to represent the properties of polymer solutions.
Illustrative results are shown in Figures 2 and 3. All re-
sults shown in Figure 3 were obtained with only two
adjustable parameters.
LIQUID-LIQUID EQUILIBRIA
Well-known equations of thermodynamic stability can
be applied to the UNIQUAC equation to determine
whether one or two liquid phases exist at equilibrium.
Whenever the magnitude of u12is low compared to that
I00 L L
- 50
I09
-
c
?!
of ull or uZ2, a single phase is unstable and two liquid
phases are formed. However, in addition to energetic
effects, differences in molecular size and shape may also
affect phase stability. To illustrate, Figure 4 shows the
molar Gibbs energy of mixing as a function of composition.
For simplicity, a symmetric system has been chosen; for
- each of the three curves shown r1 = r2 = 3.3 and (uz1 -
u
ull) = (u12 - uz2) = 0.45 RT. Further, for all three
curves q1 = 92. The lowest line, with q = 2, indicates
- phase stability; the middle line, with q = 2.5, indicates
0:
\ incipient instability and the top line, with q = 3, indicates
3
o_
- -50
0.3 I 1 \
-100 1 - 1 0 0
0 10 20 30 40 50
Temperature ("C) 0.2
Fig. 1. UNIQUAC parameters for nitroethane ( l ) , octane (2); and
nitroethane (I), hexane (3) calculated from vapor-liquid and liquid-
liquid data. The upper consolute temperature i s Tc.
E
c
0
!E3 0.
-Uniquac Equation Lo
v)
Experimental 2
-
0
E
0
a
L
0
a
-
2
L
>
0 t
'-
.n
L 1
5)
L Q)
a c
iz _ - --Uniquac Equation
L
0
3
0
J.
>
Experimental
-0
a,
c
a
l
N
0.05 I0 I I
I I J 0.0 I d
0 0.2 0.4 0.6 0
0.2 0.4 0.6 0.8 I .o
Weight Froct ion Benzene Weight Fraction Water
Fig. 3. Vapor pressure of water in solutions of polyethylene glycol a t
Fig. 2. Vapor pressure of benzene in solutions of polyisobuwlene 65C.
(molecular weight lo5) a t 40C.
AlChE Journal (Vol. 21, No. 1)
Page 122 January, 1975
that two phases coexist. Figure 4, therefore, shows that the than that of an unbranched chain having the same num-
tendency for phase-splitting depends not only on differ- ber of segments. Therefore, we expect methanol-n-octane
ences in potential energy but also on molecular geometry: to show phase instability more readily than does methanol-
when q = 2, the external molecular surface area is insuffi- iso-octane, in agreement with experiment: the upper con-
cient to produce phase instability but when q = 3 the solute temperature for methanol-n-octane is 66.7 "C while
molecular area available for interaction is sufficiently large that for methanol-iso-octane is 42.5"C (Francis, 1961).
to produce two liquid phases. When binary data alone are used, representation of
The effect illustrated in Figure 4 is observed in experi- multicomponent liquid-liquid equilibria is much more difi-
mental results for the systems methanol-n-octane and cult than representation of multicomponent vapor-liquid
methanol-iso-octane. In these two binaries, the energetic equilibria. For vapor-liquid equilibria for mixtures of nor-
parameters are essentially identical and parameter T for mal molecular weight, Raoult's law often gives a reason-
octane is almost independent of chain branching. How- able zeroth or first approximation; activity coefficients
ever, q for iso-octane is smaller than that for n-octane; the provide corrections, but the main influence on vapor-
external surface area of a branched chain is always lower liquid equilibria is given by the pure-component vapor
pressures. However, this influence plays no role at all in
liquid-liquid equilibria. Small changes in activity coeffi-
0 cients usually produce large qualitative changes in the
I I I I 1 shape of the connodal curve and the slopes of the tie
lines, especially near the plait point. Therefore, for any
empirical or semi-empirical expression for the excess Gibbs
energy, ternary liquid-liquid equilibria provide a much
stricter test than ternary vapor-liquid equilibria.
The two-parameter Wilson equation which is so suc-
cessful for multicomponent vapor-liquid equilibria, fails
completely for liquid-liquid equilibria (Wilson, 1964),
even in binary systems, While Wilson has suggested a
three-parameter form of his equation which is applicable
to binary liquid-liquid systems (Wilson, 1964), it is not
.- possible to generalize it to ternary (and higher) systems
E
-0.15 I-\\ without additional simplifying assumptions (Renon and
Prausnitz, 1969). The two-parameter equations of van
Laar and Margules are applicable to liquid-liquid equi-
libria but usually give poor representation of such equi-
libria.
The three-parameter NRTL equation, proposed in 1968,
provides a large improvement in representation of ternary
L
-
0 liquid-liquid equilibria using only binary parameters,
0 However, calculated results are sensitive to the choice of
2
-0'25- - the binary parameters a12,02, and a13, especially to 012
in plait-point systems where 1 and 2 are the partially
.I
A. miscible components. Mutual solubility data for the 1-2
binary can be used to fix two of the 1-2 binary param-
eters, but the all-important choice of "12 remains elusive.
-0.30 Empirical rules originally suggested ( Renon and Praus-
nitz, 1968) appear to be inadequate for consistently re-
Mole F r a c t i o n liable results. Therefore, as shown by Renon et al. (1971),
Fig. 4. Effect of surface-area parameter q on Gibbs energy of mixing good representation for ternary systems can only be ob-
for a symmetric system according to UNIQUAC. tained with extensive ternary data used to fix the nine
I 2
Experimental
Calculated With
Uniquac E q u a t i o n I \ Calcu la ted
With Uniquac
45 A B _ _ _ Binary Data Only
B i n a r y Data a n d K F
- 8
- 7
- 6
L
- Q)
3 0
0
- 5 - c
n
.-
- 4
c
L
ln
- 3
- 2
- I
5 t
0.6 I 1 I
0 0. I 0.2 0.3
0
0 0 I 0.2 0.3 0.4 0.5 0.6 Mole F r a c t i o n Benzene In Furfural Phase
Male Fraction Acetone In Chloroform Phose Fig. 10. Distribution coefficient of benzene and selectivity of furfural
Fig. 8. Distribution coefficient of acetone and p , selectivity of for benzene relative to isooctane.
chloroform for acetone relative to water, a t 60C.
Uniquac U s i n g Ternary
- - ET irpert e Dtaa tIa Only.
e - L i nmen
\\/ ---*--
____-----
......................................................................
______----- -----
,v V V V 2 2
Fig. 9. Liquid-liquid equilibria in the system: furfural (I), isooctane Fig. 11. Liquid-liquid equilibria in the system: ethyl acetate (1).
(21, and benzene (3) a t 25C. Concentrations are in weight fractions. water (2). and ethanol (3). Concentrations are in weight fractions.
The population balances describing the time dependence of the size dis-
tibution can, under some conditions, be transformed by means of a similar-
ity transformation into an ordinary integro-diff erential equation containing
two instead of three variables, If there is compatibility between the trans- B. PULVERMACHER
formed equation and the constraints given by the total mass conservation Department of Chemical Engineering
equation and the equation for the total number of particles, a self-preserv- University of Delaware
ing spectrum of the first kind can be obtained. There are, however, many Newark, Delaware 1971 1
situations such as the sintering controlled aging of supported metal cata-
lysts, coagulation of colloidal particles in laminar shear flow, and coagula- and
tion of colloidal particles in a turbulent flow when the particles are smaller
than the size of the smallest eddy for which, although a similarity trans- E. RUCKENSTEIN
formation is possible, the transformed equation has no solution because of
incompatibility with the above mentioned constraints. A second kind of Faculty of Engineering and Applied Sciences
State University of New York at Buffalo
self-preserving spectrum is suggested for these situations. The new variables Buffalo, New York 14214
are induced from a particular case for which an analytical result is available.
A detailed presentation of the sintering controlled aging of supported metal
catalysts is presented.